tensorcrush 847371d2de [AGC] Decode VOP3P and emit packed f16 arithmetic (first slice) (#145)
* [AGC] Decode VOP3P and emit packed f16 arithmetic (first slice)

On gfx10 the VOP3P family lives under its own 0b110011000 prefix (word0
top byte 0xCC), which the major-opcode switch currently routes to the
SMEM branch, so packed instructions were decoded as scalar memory ops
and emitted as silent no-ops. Intercept the exact 9-bit prefix ahead of
the switch, decode the five packed-f16 arithmetic opcodes with their
op_sel/op_sel_hi/neg_lo/neg_hi/clamp modifiers, and emit them as
UnpackHalf2x16 -> component-wise f32 vec2 ops -> PackHalf2x16 so no
Float16 capability is needed.

Bit layout and opcode numbers pinned to LLVM MC test encodings
(vop3p.s, gfx10_vop3p_literalv216.txt) and VOP3PInstructions.td.
Unsupported modifiers, packed constants and out-of-scope packed opcodes
fail with a clear error instead of emitting wrong results.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

* [AGC] Make packed f16 exact and drop v_pk_fma_f16 (review response)

Address the FP16 correctness review on the VOP3P slice.

Replace GLSL UnpackHalf2x16/PackHalf2x16 with explicit integer f16<->f32
conversions (EmitHalfToFloat/EmitFloatToHalf): exact widening with subnormal
normalisation, and narrowing with round-to-nearest-even, overflow-to-Inf and
NaN/Inf handling. Their subnormal and rounding behaviour no longer depends on
implementation-defined float-controls modes.

With exact conversions, v_pk_add_f16 and v_pk_mul_f16 are bit-exact to a true
f16 op (f32 result rounds losslessly to f16; a f16 product fits in f32). Emit
v_pk_min_f16/v_pk_max_f16 as fminnum_like/fmaxnum_like (NaN operand returns the
other; ordered numeric compare) instead of GLSL FMin/FMax.

v_pk_fma_f16 now fails emission loudly: a fused f16 FMA rounds once, an f32
multiply-add then pack double-rounds (fma(0x4100,0x7522,0x04EA) is 0x7A6B fused
vs 0x7A6A via f32). Exact fused emulation is a planned follow-up slice.

ShaderDump gains an Expect model (Translates/DecodeFails/EmitFails) and packed
regressions: arith, non-default modifiers, and loud-failure pins for the fma
case above and for clamp.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: tensorcrush <tensorcrush@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Co-authored-by: Antigravity AI <antigravity@gemini.com>
2026-07-17 18:58:22 +03:00
2026-03-11 15:48:28 +03:00
2026-07-17 02:47:53 +03:00

SharpEmu

An experimental PlayStation 5 emulator for Windows, Linux and macOS.

Join our Discord

Join our Discord for development updates, compatibility discussions, support, and community chat.


Note

SharpEmu supports Windows x64, Linux x64, and macOS x64. Apple Silicon Macs can run the macOS x64 build through Rosetta 2.

Warning

SharpEmu is an experimental PS5 emulator developed from scratch in C#. The current focus is on accuracy and infrastructure setup rather than game-specific compatibility.

Info

SharpEmu is an emulator project currently in its early stages of development.

This project is developed purely for research and educational purposes. There are no commercial goals associated with it. We enjoy learning about system architecture and reverse engineering.

SharpEmu focuses exclusively on the PlayStation 5.
Our goal is not to emulate PS4 games, as there is already an excellent emulator dedicated to that platform: ShadPS4.

Games Tested

Demons Souls Remake Dreaming Sarah
Bloodborne screenshot Dreaming Sarah
Void Terrarium Dead Cells
Void Terrarium Dead Cells

Status

The emulator can currently load the eboot.bin of real games, execute native CPU instructions, and partially handle kernel-related functionality. However, several critical components are still missing.

Current capabilities include:

  • Loading eboot.bin and .elf files
  • Executing native CPU instructions
  • Reading basic game metadata (title, version, etc.)
  • Loading system modules (prx / sys_module)
  • Partial support for some kernel functions
  • Fiber and AMPR exports
  • PlayGo scenarios
  • Initial loading game files
  • Shader/resource submits and AGC initial
  • Video outputs in some games

Some games have reached like sceVideoOut and AGC stages.

SharpEmu supports Windows, Linux, and macOS hosts. Video output uses Vulkan on Windows and Linux, and MoltenVK on macOS. Platform support is still experimental, so compatibility and performance vary by game, operating system, and GPU driver.

Using

Download the release archive for your operating system, extract it, and launch SharpEmu with the path to a legally obtained game's eboot.bin.

Windows PowerShell:

.\SharpEmu.exe "C:\path\to\game\eboot.bin" 2>&1 |
  Tee-Object -FilePath "SharpEmu.log"

Linux and macOS:

chmod +x ./SharpEmu

./SharpEmu "/path/to/game/eboot.bin" 2>&1 |
  tee SharpEmu.log

A Vulkan-capable GPU and current graphics driver are required. The macOS release includes the MoltenVK Vulkan implementation.

Important

This project does not support or condone piracy.
All games used during development and testing are dumped from consoles that we personally own.
Users are expected to use legally obtained copies of their games.

Build

  1. Install the .NET SDK version specified in global.json.
  2. Clone the repository: git clone https://github.com/sharpemu/sharpemu.git
  3. Open the solution file (SharpEmu.slnx) in VSCode.
  4. Build the project: dotnet build or dotnet publish
  5. Build artifacts will be located in the artifacts directory.

Disclaimer

SharpEmu is an experimental emulator intended for research and educational purposes.

This project does not contain any copyrighted system firmware, game data, or proprietary PlayStation assets.

Special Thanks

The following projects were extremely helpful during development:

  • ShadPS4
    Helped with understanding the basic architecture of the PlayStation 4.

  • Kyty
    One of the few PS5 emulator projects available and very useful for studying native code execution.

  • Ryujinx
    Provided valuable references for filesystem handling and low-level C# implementation patterns.

License

Contributing

Before opening an issue or pull request, please read our contribution guidelines:

CONTRIBUTING.md

The guide covers:

  • Coding style and formatting
  • AI-assisted contributions
  • Pull request expectations
  • Testing guidelines
  • Legal and reverse engineering policy
S
Description
Languages
C# 97.6%
Python 1.3%
JavaScript 0.4%
CSS 0.4%
HTML 0.2%